ml-SVR

ml-SVR identifies gene regulatory modules by applying a multilevel support vector regression framework that integrates binding motif information with gene expression data to suppress false positives in analyses of yeast cell cycle and human cancer datasets, including breast cancer estrogen-response studies.

Key Features:

• Multilevel Analysis Strategy: Progressively refines gene sets across multiple levels to increase the significance and reliability of identified regulatory modules.
• Two-Stage Support Vector Regression (SVR): Applies a two-stage SVR procedure at each analysis level to model regulatory relationships.
• Integration of Binding Motif Information and Gene Expression: Combines motif-derived regulatory evidence with gene expression data to inform SVR-based inference.
• False Positive Suppression: Explicitly reduces false positive regulatory predictions through multilevel refinement and two-stage SVR.
• Significance Assessment: Performs statistical significance analysis for each candidate regulatory module to validate biological relevance.

Scientific Applications:

• Benchmarking with simulated and yeast cell cycle data: Demonstrated comparative analyses showing improved accuracy in identifying regulators and target genes relative to existing methods on simulation data and yeast cell cycle datasets.
• Breast cancer estrogen-response analysis: Identified condition-specific regulatory modules associated with estrogen treatment in breast cancer cell line data, providing insights into estrogen signaling and action.
• Application to complex human cancer datasets: Suited for regulatory inference in complex datasets such as human cancer studies where false positive control is critical.

Methodology:

Multilevel analysis that progressively refines gene sets; two-stage SVR applied at each level integrating binding motif information with gene expression data; statistical significance analysis of each candidate regulatory module.